A hierarchical multiclass support vector machine incorporated with holistic triple learning units

This paper proposes a new hierarchical learning structure, namely the holistic triple learning (HTL), for extending the binary support vector machine (SVM) to multi-classification problems. For an N-class problem, a HTL constructs a decision tree up to a depth of A leaf node of the decision tree is allowed to be placed with a holistic triple learning unit whose generalisation abilities are assessed and approved. Meanwhile, the remaining nodes in the decision tree each accommodate a standard binary SVM classifier. The holistic triple classifier is a regression model trained on three classes, whose training algorithm is originated from a recently proposed implementation technique, namely the least-squares support vector machine (LS-SVM). A major novelty with the holistic triple classifier is the reduced number of support vectors in the solution. For the resultant HTL-SVM, an upper bound of the generalisation error can be obtained. The time complexity of training the HTL-SVM is analysed, and is shown to be comparable to that of training the one-versus-one (1-vs.-1) SVM, particularly on small-scale datasets. Empirical studies show that the proposed HTL-SVM achieves competitive classification accuracy with a reduced number of support vectors compared to the popular 1-vs-1 alternative.

Spitzer infrared spectrograph point source classification in the Small Magellanic Cloud

The Magellanic Clouds are uniquely placed to study the stellar contribution to dust emission. Individual stars can be resolved in these systems even in the mid-infrared, and they are close enough to allow detection of infrared excess caused by dust. We have searched the Spitzer Space Telescope data archive for all Infrared Spectrograph (IRS) staring-mode observations of the Small Magellanic Cloud (SMC) and found that 209 Infrared Array Camera (IRAC) point sources within the footprint of the Surveying the Agents of Galaxy Evolution in the Small Magellanic Cloud (SAGE-SMC) Spitzer Legacy programme were targeted, within a total of 311 staring-mode observations. We classify these point sources using a decision tree method of object classification, based on infrared spectral features, continuum and spectral energy distribution shape, bolometric luminosity, cluster membership and variability information. We find 58 asymptotic giant branch (AGB) stars, 51 young stellar objects, 4 post-AGB objects, 22 red supergiants, 27 stars (of which 23 are dusty OB stars), 24 planetary nebulae (PNe), 10 Wolf-Rayet stars, 3 H II regions, 3 R Coronae Borealis stars, 1 Blue Supergiant and 6 other objects, including 2 foreground AGB stars. We use these classifications to evaluate the success of photometric classification methods reported in the literature.

Interpretable and Reconfigurable Clustering of Document Datasets by Deriving Word-based Rules

Clusters of text documents output by clustering algorithms are often hard to interpret. We describe motivating real-world scenarios that necessitate reconfigurability and high interpretability of clusters and outline the problem of generating clusterings with interpretable and reconfigurable cluster models. We develop two clustering algorithms toward the outlined goal of building interpretable and reconfigurable cluster models. They generate clusters with associated rules that are composed of conditions on word occurrences or nonoccurrences. The proposed approaches vary in the complexity of the format of the rules; RGC employs disjunctions and conjunctions in rule generation whereas RGC-D rules are simple disjunctions of conditions signifying presence of various words. In both the cases, each cluster is comprised of precisely the set of documents that satisfy the corresponding rule. Rules of the latter kind are easy to interpret, whereas the former leads to more accurate clustering. We show that our approaches outperform the unsupervised decision tree approach for rule-generating clustering and also an approach we provide for generating interpretable models for general clusterings, both by significant margins. We empirically show that the purity and f-measure losses to achieve interpretability can be as little as 3 and 5%, respectively using the algorithms presented herein.

Rastreio do cancro da mama no concelho de Aveiro: do estudo dos fatores determinantes da adesão às propostas educativas emergentes

O cancro da mama feminino pela sua magnitude merece uma especial atenção ao nível das políticas de saúde. Emerge, pois uma visão abrangente que, por um lado, deve atentar para o encargo que esta representa para qualquer sistema de saúde, pelos custos que acarreta, como também, para a qualidade de vida das mulheres portadoras da mesma. Desta forma, a Liga Portuguesa Contra o Cancro (LPCC) tem desenvolvido, em colaboração com as Administrações Regionais de Saúde (ARS), o Programa de Rastreio do Cancro da Mama (PRCM), o qual apresenta, no Concelho de Aveiro, taxas de adesão na ordem dos 50%, ainda distantes dos 70%, objetivo recomendado pelas guidelines da Comissão Europeia. A não adesão tem sido considerada como um dos principais problemas do sistema de saúde, tanto pelas repercussões ao nível de ganhos em saúde, como também na qualidade de vida e na satisfação dos pacientes com os cuidados de saúde, constituindo-se como um fenómeno multifatorial e multidimensional. É neste sentido que o presente trabalho se propõe identificar os fatores, de cariz individual e do meio envolvente, determinantes da adesão ao PRCM, numa amostra de mulheres residentes no Concelho de Aveiro, com idades compreendidas entre os 45 e os 69 anos e, a partir dos resultados emergentes, propor estratégias de educação em saúde. Como procedimentos metodológicos e, numa primeira fase, entre outubro 2009 e maio 2010 foi aplicado um survey, o qual foi complementado com notas de campo dos entrevistadores a uma amostra não aleatória de 805 mulheres, em dois contextos distintos: no centro de saúde às aderentes à mamografia e, no domicílio, às não aderentes. Numa segunda fase, realizamos duas sessões de Focus Group (FG), num total de 12 elementos, um grupo heterogéneo com enfermeiros, médicos e utentes, e um outro grupo homogéneo, apenas com profissionais de saúde. O tratamento dos dados do survey foi efetuado através de procedimentos estatísticos, com utilização do SPSS® versão 17 e realizadas análises bivariadas (qui-quadrado) e multivariadas (discriminação de função e árvore de decisão através do algoritmo Chi-squared Automatic Interaction Detector) com o intuito de determinar as diferenças entre os grupos e predizer as variáveis exógenas. No que diz respeito a indicadores sociodemográficos, os resultados mostram que aderem mais, as mulheres com idades <50 anos e ≥ 56 anos, as que vivem em localidades urbanas, as trabalhadoras não qualificadas e as reformadas. As que aderem menos ao PRCM têm idades compreendidas entre os 50-55 anos, vivem nas zonas periurbanas, são licenciadas, apresentam categoria profissional superior ou estão desempregadas. Em relação às restantes variáveis exógenas, aderem ao PRCM, as mulheres que apresentam um Bom Perfil de Conhecimentos (46.6%), enquanto as não aderentes apresentam um Fraco Perfil de Conhecimentos (50.6%), sendo esta relação estatisticamente significativa (X2= 10.260; p=0.006).Cerca de 59% das mulheres aderentes realiza o seu rastreio de forma concordante com as orientações programáticas presentes no PRCM, comparativamente com 41.1% das mulheres que não o faz, verificando-se uma relação de dependência bastante significativa entre as variáveis Perfil de Comportamentos e adesão(X2= 348.193; p=0.000). Apesar de não existir dependência estatisticamente significativa entre as Motivações e a adesão ao PRCM (X2= 0.199; p=0.656), se analisarmos particularmente, os motivos de adesão, algumas inquiridas demonstram preocupação, tanto na deteção precoce da doença, como na hereditariedade. Por outro lado, os motivos de não adesão, também denotam aspetos de nível pessoal como o desleixo com a saúde, o desconhecimento e o esquecimento da marcação. As mulheres que revelam Boa Acessibilidade aos Cuidados de Saúde Primários e um Bom Atendimento dos Prestadores de Cuidados aderem mais ao PRCM, comparativamente com as inquiridas que relatam Fraca Acessibilidade e Atendimento, não aderindo. A partir dos resultados da análise multivariada podemos inferir que as variáveis exógenas estudadas possuem um poder discriminante significativo, sendo que, o Perfil de Comportamentos é a variável que apresenta maior grau de diferenciação entre os grupos das aderentes e não aderentes. Como variáveis explicativas resultantes da árvore de decisão CHAID, permaneceram, o Perfil de Comportamentos (concordantes e não concordantes com as guidelines), os grupos etários (<50 anos, 50-55anos e ≥56anos) e o Atendimento dos prestadores de cuidados de saúde. As mulheres mais novas (<50 anos) com Perfil de comportamentos «concordantes» com as guidelines são as que aderem mais, comparativamente com os outros grupos etários. Por outro lado, as não aderentes necessitam de um «bom» atendimento dos prestadores de cuidados para se tornarem aderentes ao PRCM. Tanto as notas de campo, como a discussão dos FG foram sujeitas a análise de conteúdo segundo as categorias em estudo obtidas na primeira fase e os relatos mostram a importância de fatores de ordem individual e do meio envolvente. No que se refere a aspetos psicossociais, destaca-se a importância das crenças e como fatores ambientais menos facilitadores para a adesão apontam a falta de transportes, a falta de tempo das pessoas e a oferta de recursos, principalmente se existirem radiologistas privados como alternativa ao PRCM. Tal como na primeira fase do estudo, uma das motivações para a adesão é a recomendação dos profissionais de saúde para o PRCM, bem como a marcação de consultas pela enfermeira, que pode ser uma oportunidade de contacto para a sensibilização. Os hábitos de vigilância de saúde, a perceção positiva acerca dos programas de saúde no geral, o acesso à informação pertinente sobre o PRCM e a operacionalização deste no terreno parecem ser fatores determinantes segundo a opinião dos elementos dos FG. O tipo e a regularidade no atendimento por parte dos profissionais de saúde, a relação entre profissional de saúde/paciente, a personalização das intervenções educativas, a divulgação que estes fazem do PRCM junto das suas pacientes, bem como, a organização do modelos de cuidados de saúde das unidades de saúde e a forma como os profissionais se envolvem e tomam a responsabilização por um programa desta natureza são fatores condicionantes da adesão. Se atendermos aos resultados deste estudo, verificamos um envolvimento de fatores que integram múltiplos níveis de intervenção, sendo um desafio para as equipas de saúde que pretendam intervir no âmbito do programa de rastreio do cancro da mama. Com efeito, os resultados também apontam para a combinação de múltiplas estratégias que são transversais a vários programas de promoção da saúde, assumindo, desta forma, uma perspetiva multidimensional e dinâmica que visa, essencialmente, a construção social da saúde e do bem-estar (i.e. responsabilização do cidadão pela sua própria saúde e o seu empowerment).

Dimensionamento de obras de dissipação de energia em drenagem de vias de comunicação

Dissertação de Natureza Científica para obtenção do grau de Mestre em Engenharia Civil na Área de Especialização de Hidráulica

Economic Evaluation of Ticagrelor for Secondary Prevention Following Acute Coronary Syndromes

INTRODUCTION AND OBJECTIVES: To estimate the cost-effectiveness and cost-utility of ticagrelor in the treatment of patients with acute coronary syndromes (unstable angina or myocardial infarction with or without ST-segment elevation), including patients treated medically and those undergoing percutaneous coronary intervention or coronary artery bypass grafting. METHODS: A short-term decision tree and a long-term Markov model were used to simulate the evolution of patients' life-cycles. Clinical effectiveness data were collected from the PLATO trial and resource use data were obtained from the Hospital de Santa Marta database, disease-related group legislation and the literature. RESULTS: Ticagrelor provides increases of 0.1276 life years and 0.1106 quality-adjusted life years (QALYs) per patient. From a societal perspective these clinical gains entail an increase in expenditure of €610. Thus the incremental cost per life year saved is €4780 and the incremental cost per QALY is €5517. CONCLUSIONS: The simulation results show that ticagrelor reduces events compared to clopidogrel. The costs of ticagrelor are partially offset by lower costs arising from events prevented. The use of ticagrelor in clinical practice is therefore cost-effective compared to generic clopidogrel.

American Society of Anesthesiologists Score: Still Useful After 60 Years? Results of the EuSOS Study

OBJECTIVE: The European Surgical Outcomes Study described mortality following in-patient surgery. Several factors were identified that were able to predict poor outcomes in a multivariate analysis. These included age, procedure urgency, severity and type and the American Association of Anaesthesia score. This study describes in greater detail the relationship between the American Association of Anaesthesia score and postoperative mortality. METHODS: Patients in this 7-day cohort study were enrolled in April 2011. Consecutive patients aged 16 years and older undergoing inpatient non-cardiac surgery with a recorded American Association of Anaesthesia score in 498 hospitals across 28 European nations were included and followed up for a maximum of 60 days. The primary endpoint was in-hospital mortality. Decision tree analysis with the CHAID (SPSS) system was used to delineate nodes associated with mortality. RESULTS: The study enrolled 46,539 patients. Due to missing values, 873 patients were excluded, resulting in the analysis of 45,666 patients. Increasing American Association of Anaesthesia scores were associated with increased admission rates to intensive care and higher mortality rates. Despite a progressive relationship with mortality, discrimination was poor, with an area under the ROC curve of 0.658 (95% CI 0.642 - 0.6775). Using regression trees (CHAID), we identified four discrete American Association of Anaesthesia nodes associated with mortality, with American Association of Anaesthesia 1 and American Association of Anaesthesia 2 compressed into the same node. CONCLUSION: The American Association of Anaesthesia score can be used to determine higher risk groups of surgical patients, but clinicians cannot use the score to discriminate between grades 1 and 2. Overall, the discriminatory power of the model was less than acceptable for widespread use.

Molecular tools in the diagnostic and epidemiology of infections caused by members of Mycobacterium avium Complex

Mycobacterium avium Complex (MAC) comprises microorganisms that affect a wide range of animals including humans. The most relevant are Mycobacterium avium subspecies hominissuis (Mah) with a high impact on public health affecting mainly immunocompromised individuals and Mycobacterium avium subspecies paratuberculosis (Map) causing paratuberculosis in animals with a high economic impact worldwide. In this work, we characterized 28 human and 67 porcine Mah isolates and evaluated the relationship among them by Multiple-Locus Variable number tandem repeat Analysis (MLVA). We concluded that Mah population presented a high genetic diversity and no correlations were inferred based on geographical origin, host or biological sample. For the first time in Portugal Map strains, from asymptomatic bovine faecal samples were isolated highlighting the need of more reliable and rapid diagnostic methods for Map direct detection. Therefore, we developed an IS900 nested real time PCR with high sensitivity and specificity associated with optimized DNA extraction methodologies for faecal and milk samples. We detected 83% of 155 faecal samples from goats, cattle and sheep, and 26% of 98 milk samples from cattle, positive for Map IS900 nested real time PCR. A novel SNPs (single nucleotide polymorphisms) assay to Map characterization based on a Whole Genome Sequencing analysis was developed to elucidate the genetic relationship between strains. Based on sequential detection of 14 SNPs and on a decision tree we were able to differentiate 14 phylogenetic groups with a higher discriminatory power compared to other typing methods. A pigmented Map strain was isolated and characterized evidencing for the first time to our knowledge the existence of pigmented Type C strains. With this work, we intended to improve the ante mortem direct molecular detection of Map, to conscientiously aware for the existence of Map animal infections widespread in Portugal and to contribute to the improvement of Map and Mah epidemiological studies.

Mapping the Pasture Steppe in Bayankhongor, Mongolia: comparison of classification methods, using Landsat-8 and geophysical data

Grasslands in semi-arid regions, like Mongolian steppes, are facing desertification and degradation processes, due to climate change. Mongolia’s main economic activity consists on an extensive livestock production and, therefore, it is a concerning matter for the decision makers. Remote sensing and Geographic Information Systems provide the tools for advanced ecosystem management and have been widely used for monitoring and management of pasture resources. This study investigates which is the higher thematic detail that is possible to achieve through remote sensing, to map the steppe vegetation, using medium resolution earth observation imagery in three districts (soums) of Mongolia: Dzag, Buutsagaan and Khureemaral. After considering different thematic levels of detail for classifying the steppe vegetation, the existent pasture types within the steppe were chosen to be mapped. In order to investigate which combination of data sets yields the best results and which classification algorithm is more suitable for incorporating these data sets, a comparison between different classification methods were tested for the study area. Sixteen classifications were performed using different combinations of estimators, Landsat-8 (spectral bands and Landsat-8 NDVI-derived) and geophysical data (elevation, mean annual precipitation and mean annual temperature) using two classification algorithms, maximum likelihood and decision tree. Results showed that the best performing model was the one that incorporated Landsat-8 bands with mean annual precipitation and mean annual temperature (Model 13), using the decision tree. For maximum likelihood, the model that incorporated Landsat-8 bands with mean annual precipitation (Model 5) and the one that incorporated Landsat-8 bands with mean annual precipitation and mean annual temperature (Model 13), achieved the higher accuracies for this algorithm. The decision tree models consistently outperformed the maximum likelihood ones.

IFN stimulated gene expression in the liver is a better predictor of treatment response in chronic hepatitis c than the IL28b (IFN lambda 3) genotype

Background: Therapy of chronic hepatitis C (CHC) with pegIFNa/ribavirin achieves sustained virologic response (SVR) in ~55%. Pre-activation of the endogenous interferon system in the liver is associated non-response (NR). Recently, genome-wide association studies described associations of allelic variants near the IL28B (IFNλ3) gene with treatment response and with spontaneous clearance of the virus. We investigated if the IL28B genotype determines the constitutive expression of IFN stimulated genes (ISGs) in the liver of patients with CHC. Methods: We genotyped 93 patients with CHC for 3 IL28B single nucleotide polymorphisms (SNPs, rs12979860, rs8099917, rs12980275), extracted RNA from their liver biopsies and quantified the expression of IL28B and of 8 previously identified classifier genes which discriminate between SVR and NR (IFI44L, RSAD2, ISG15, IFI22, LAMP3, OAS3, LGALS3BP and HTATIP2). Decision tree ensembles in the form of a random forest classifier were used to calculate the relative predictive power of these different variables in a multivariate analysis. Results: The minor IL28B allele (bad risk for treatment response) was significantly associated with increased expression of ISGs, and, unexpectedly, with decreased expression of IL28B. Stratification of the patients into SVR and NR revealed that ISG expression was conditionally independent from the IL28B genotype, i.e. there was an increased expression of ISGs in NR compared to SVR irrespective of the IL28B genotype. The random forest feature score (RFFS) identified IFI27 (RFFS = 2.93), RSAD2 (1.88) and HTATIP2 (1.50) expression and the HCV genotype (1.62) as the strongest predictors of treatment response. ROC curves of the IL28B SNPs showed an AUC of 0.66 with an error rate (ERR) of 0.38. A classifier with the 3 best classifying genes showed an excellent test performance with an AUC of 0.94 and ERR of 0.15. The addition of IL28B genotype information did not improve the predictive power of the 3-gene classifier. Conclusions: IL28B genotype and hepatic ISG expression are conditionally independent predictors of treatment response in CHC. There is no direct link between altered IFNλ3 expression and pre-activation of the endogenous system in the liver. Hepatic ISG expression is by far the better predictor for treatment response than IL28B genotype.

Apports de la segmentation de la population des dirigeants au processus de gestion des ressources humaines : Le cas de l’administration algérienne

A partir des résultats d’une enquête effectuée en 2005 sur un échantillon de 203 dirigeants publics, une typologie floue de trois profils a été dégagée en vue de concevoir un système d’affectation des dirigeants en fonction de leur style du leadership, sens du travail, et leurs préoccupations de gestion des ressources humaines. En se basant sur cette typologie floue, des techniques empruntées à l’intelligence artificielle ont été appliquées pour apprendre des règles de classification. Ces techniques sont au nombre de quatre : le réseau neuronal (Neural Network), l’algorithme génétique (Genetic Algorithm), l’arbre de décision (Decision Tree) et la théorie des ensembles approximatifs (Rough Sets). Les résultats de l’étude ainsi que ses perspectives seront présentées et discutés tout au long de cette communication.

Economic evaluation of benzodiazepines versus cognitive behavioural therapy among older adults with chronic insomnia

L’insomnie, commune auprès de la population gériatrique, est typiquement traitée avec des benzodiazépines qui peuvent augmenter le risque des chutes. La thérapie cognitive-comportementale (TCC) est une intervention non-pharmacologique ayant une efficacité équivalente et aucun effet secondaire. Dans la présente thèse, le coût des benzodiazépines (BZD) sera comparé à celui de la TCC dans le traitement de l’insomnie auprès d’une population âgée, avec et sans considération du coût additionnel engendré par les chutes reliées à la prise des BZD. Un modèle d’arbre décisionnel a été conçu et appliqué selon la perspective du système de santé sur une période d’un an. Les probabilités de chutes, de visites à l’urgence, d’hospitalisation avec et sans fracture de la hanche, les données sur les coûts et sur les utilités ont été recueillies à partir d’une revue de la littérature. Des analyses sur le coût des conséquences, sur le coût-utilité et sur les économies potentielles ont été faites. Des analyses de sensibilité probabilistes et déterministes ont permis de prendre en considération les estimations des données. Le traitement par BZD coûte 30% fois moins cher que TCC si les coûts reliés aux chutes ne sont pas considérés (231$ CAN vs 335$ CAN/personne/année). Lorsque le coût relié aux chutes est pris en compte, la TCC s’avère être l’option la moins chère (177$ CAN d’économie absolue/ personne/année, 1,357$ CAN avec les BZD vs 1,180$ pour la TCC). La TCC a dominé l’utilisation des BZD avec une économie moyenne de 25, 743$ CAN par QALY à cause des chutes moins nombreuses observées avec la TCC. Les résultats des analyses d’économies d’argent suggèrent que si la TCC remplaçait le traitement par BZD, l’économie annuelle directe pour le traitement de l’insomnie serait de 441 millions de dollars CAN avec une économie cumulative de 112 billions de dollars canadiens sur une période de cinq ans. D’après le rapport sensibilité, le traitement par BZD coûte en moyenne 1,305$ CAN, écart type 598$ (étendue : 245-2,625)/personne/année alors qu’il en coûte moyenne 1,129$ CAN, écart type 514$ (étendue : 342-2,526)/personne/année avec la TCC. Les options actuelles de remboursement de traitements pharmacologiques au lieu des traitements non-pharmacologiques pour l’insomnie chez les personnes âgées ne permettent pas d’économie de coûts et ne sont pas recommandables éthiquement dans une perspective du système de santé.

The use of artificial intelligence algorithms to guide surgical treatment of adolescent idiopathic scoliosis

La scoliose idiopathique de l’adolescent (SIA) est une déformation tri-dimensionelle du rachis. Son traitement comprend l’observation, l’utilisation de corsets pour limiter sa progression ou la chirurgie pour corriger la déformation squelettique et cesser sa progression. Le traitement chirurgical reste controversé au niveau des indications, mais aussi de la chirurgie à entreprendre. Malgré la présence de classifications pour guider le traitement de la SIA, une variabilité dans la stratégie opératoire intra et inter-observateur a été décrite dans la littérature. Cette variabilité s’accentue d’autant plus avec l’évolution des techniques chirurgicales et de l’instrumentation disponible. L’avancement de la technologie et son intégration dans le milieu médical a mené à l’utilisation d’algorithmes d’intelligence artificielle informatiques pour aider la classification et l’évaluation tridimensionnelle de la scoliose. Certains algorithmes ont démontré être efficace pour diminuer la variabilité dans la classification de la scoliose et pour guider le traitement. L’objectif général de cette thèse est de développer une application utilisant des outils d’intelligence artificielle pour intégrer les données d’un nouveau patient et les évidences disponibles dans la littérature pour guider le traitement chirurgical de la SIA. Pour cela une revue de la littérature sur les applications existantes dans l’évaluation de la SIA fut entreprise pour rassembler les éléments qui permettraient la mise en place d’une application efficace et acceptée dans le milieu clinique. Cette revue de la littérature nous a permis de réaliser que l’existence de “black box” dans les applications développées est une limitation pour l’intégration clinique ou la justification basée sur les évidence est essentielle. Dans une première étude nous avons développé un arbre décisionnel de classification de la scoliose idiopathique basé sur la classification de Lenke qui est la plus communément utilisée de nos jours mais a été critiquée pour sa complexité et la variabilité inter et intra-observateur. Cet arbre décisionnel a démontré qu’il permet d’augmenter la précision de classification proportionnellement au temps passé à classifier et ce indépendamment du niveau de connaissance sur la SIA. Dans une deuxième étude, un algorithme de stratégies chirurgicales basé sur des règles extraites de la littérature a été développé pour guider les chirurgiens dans la sélection de l’approche et les niveaux de fusion pour la SIA. Lorsque cet algorithme est appliqué à une large base de donnée de 1556 cas de SIA, il est capable de proposer une stratégie opératoire similaire à celle d’un chirurgien expert dans prêt de 70% des cas. Cette étude a confirmé la possibilité d’extraire des stratégies opératoires valides à l’aide d’un arbre décisionnel utilisant des règles extraites de la littérature. Dans une troisième étude, la classification de 1776 patients avec la SIA à l’aide d’une carte de Kohonen, un type de réseaux de neurone a permis de démontrer qu’il existe des scoliose typiques (scoliose à courbes uniques ou double thoracique) pour lesquelles la variabilité dans le traitement chirurgical varie peu des recommandations par la classification de Lenke tandis que les scolioses a courbes multiples ou tangentielles à deux groupes de courbes typiques étaient celles avec le plus de variation dans la stratégie opératoire. Finalement, une plateforme logicielle a été développée intégrant chacune des études ci-dessus. Cette interface logicielle permet l’entrée de données radiologiques pour un patient scoliotique, classifie la SIA à l’aide de l’arbre décisionnel de classification et suggère une approche chirurgicale basée sur l’arbre décisionnel de stratégies opératoires. Une analyse de la correction post-opératoire obtenue démontre une tendance, bien que non-statistiquement significative, à une meilleure balance chez les patients opérés suivant la stratégie recommandée par la plateforme logicielle que ceux aillant un traitement différent. Les études exposées dans cette thèse soulignent que l’utilisation d’algorithmes d’intelligence artificielle dans la classification et l’élaboration de stratégies opératoires de la SIA peuvent être intégrées dans une plateforme logicielle et pourraient assister les chirurgiens dans leur planification préopératoire.

Distributed conditional computation

L'objectif de cette thèse est de présenter différentes applications du programme de recherche de calcul conditionnel distribué. On espère que ces applications, ainsi que la théorie présentée ici, mènera à une solution générale du problème d'intelligence artificielle, en particulier en ce qui a trait à la nécessité d'efficience. La vision du calcul conditionnel distribué consiste à accélérer l'évaluation et l'entraînement de modèles profonds, ce qui est très différent de l'objectif usuel d'améliorer sa capacité de généralisation et d'optimisation. Le travail présenté ici a des liens étroits avec les modèles de type mélange d'experts. Dans le chapitre 2, nous présentons un nouvel algorithme d'apprentissage profond qui utilise une forme simple d'apprentissage par renforcement sur un modèle d'arbre de décisions à base de réseau de neurones. Nous démontrons la nécessité d'une contrainte d'équilibre pour maintenir la distribution d'exemples aux experts uniforme et empêcher les monopoles. Pour rendre le calcul efficient, l'entrainement et l'évaluation sont contraints à être éparse en utilisant un routeur échantillonnant des experts d'une distribution multinomiale étant donné un exemple. Dans le chapitre 3, nous présentons un nouveau modèle profond constitué d'une représentation éparse divisée en segments d'experts. Un modèle de langue à base de réseau de neurones est construit à partir des transformations éparses entre ces segments. L'opération éparse par bloc est implémentée pour utilisation sur des cartes graphiques. Sa vitesse est comparée à deux opérations denses du même calibre pour démontrer le gain réel de calcul qui peut être obtenu. Un modèle profond utilisant des opérations éparses contrôlées par un routeur distinct des experts est entraîné sur un ensemble de données d'un milliard de mots. Un nouvel algorithme de partitionnement de données est appliqué sur un ensemble de mots pour hiérarchiser la couche de sortie d'un modèle de langage, la rendant ainsi beaucoup plus efficiente. Le travail présenté dans cette thèse est au centre de la vision de calcul conditionnel distribué émis par Yoshua Bengio. Elle tente d'appliquer la recherche dans le domaine des mélanges d'experts aux modèles profonds pour améliorer leur vitesse ainsi que leur capacité d'optimisation. Nous croyons que la théorie et les expériences de cette thèse sont une étape importante sur la voie du calcul conditionnel distribué car elle cadre bien le problème, surtout en ce qui concerne la compétitivité des systèmes d'experts.

Classification of Handwritten Characters: Definition, Identification and Utilisation of Regional Characteristics for Lower Case English Characters

A new procedure for the classification of lower case English language characters is presented in this work . The character image is binarised and the binary image is further grouped into sixteen smaller areas ,called Cells . Each cell is assigned a name depending upon the contour present in the cell and occupancy of the image contour in the cell. A data reduction procedure called Filtering is adopted to eliminate undesirable redundant information for reducing complexity during further processing steps . The filtered data is fed into a primitive extractor where extraction of primitives is done . Syntactic methods are employed for the classification of the character . A decision tree is used for the interaction of the various components in the scheme . 1ike the primitive extraction and character recognition. A character is recognized by the primitive by primitive construction of its description . Openended inventories are used for including variants of the characters and also adding new members to the general class . Computer implementation of the proposal is discussed at the end using handwritten character samples . Results are analyzed and suggestions for future studies are made. The advantages of the proposal are discussed in detail .